RDBMS Doomed?

February 13, 2009

Here is a great article from Read Write Web that discusses the place and uses for a Key/Value database such as SSDS or SimpleDB or Google App Engine.  Jason Massie argues that the DBA profession will level off in 5-7 years and then start to fall off in 10 because of the prevelence of cloud based Key/Value based databases.

During the last couple of releases of SQL Server, Microsoft has been focused on BI.  Why?  Everyone has lots of data.  Everyone knows that there is a lot of value in that data, but they need to get at it.  BI is one of the top agenda items for CIOs according to Gartner.

What does this have to do with SSDS and SimpleDB?

SSDS and SimpleDB have basically no reporting or analytics capabilities.  Microsoft has mentioned that the future for SSDS includes OLAP like capabilities.  Even if this was available today and shipped today, would companies be ready to move their entire data ecosystem to Microsoft, Amazon, or Google, re-write everything out of RDBMS and into Key/Value programming, then re-work ETL and data warhouse infrasturcture to work on a non-existant (so far, Microsoft says) architecture?

They sure aren’t going to move all of the OLTP systems they have to the cloud right now because the data is then trapped in a no-reporting storage mechanism.  Google App Engine only allows 1000 result objects per query.  Amazon only allows 5 second duration queries.  Can companies’ appetite for analytics be satiated by that?  I don’t think so.

This seems a very log way off.  Even if one of these Key/Value databases HAD reporting and analytics, it seems a minimum of 5 to 7 year for any sizable amount of migration to take place.  And this would be if there WERE analytics and reporting.  There aren’t.

Many companies have large amounts of data and money in their VSAM/COBOL systems that are still running fine that everyone said would have to have been replaced a decade ago.  Those systems aren’t going anywhere for quite a while and these systems need analytics.  These analytics will need to be done on-sight on RDBMS/OLAP platforms.

I suppose for a San Francisco-based startup-focused DBA, there might be a noticible movement to the cloud, but for the rest of the world, it is going to take longer than 5 years to notice a plateau in the need for RDBMS DBAs.

I will however conceede that at some point, there will be a draw to the cloud.  But there are many hurdles to overcome before that can happen in earnest.

One change might be a cloud hosted RDBMS.  Certainly there are some out there somewhere, but the hot topic right now is Key/Value and how that enables large distributed systems.  For a company looking to save money, a hosted RDBMS makes sense, but probably not SSDS and Key/Value.

Advertisements